CN107948659B - HEVC-SCC Adaptive Energy Consumption Control Strategy for Game Video Content - Google Patents

Abstract

The invention discloses an HEVC-SCC self-adaptive energy consumption control strategy aiming at game video content, which is characterized in that a game video sequence image is divided into 8x8 blocks, and the corresponding output value of one block is set to be Jud after a series of preprocessing_bAll of Jud_bForming a discrimination matrix, and fusing Jud elements in the discrimination matrix according to the size of the input CU_bCalculating the flatness degree of the CU; and (4) judging according to the flatness degree, and judging a region with complex color (similar to a natural image) and a region with very flat color. According to the HEVC-SCC adaptive energy consumption control strategy for the game video content, the coding rate of SC image coding can be improved, and the coding time is shortened on the premise that the coding quality is not influenced too much.

Description

HEVC-SCC Adaptive Energy Consumption Control Strategy for Game Video Content

technical field

The present invention relates to the technical field of video coding, in particular to a power consumption control strategy for a coding standard of Screen Content (extended SCC of HEVC).

Background technique

SC images specifically refer to images generated by computers. In recent years, with the continuous development of video content, SC images have become more and more deeply embedded in people's lives. In January 2014, the MPEG Requirements subgroup published requirements for screen contentcoding (ie Call for Proposals, CfP), and finally published the first draft of HEVC ScreenContent Coding (HEVC-SCC) in July 2014. Mainly, four new coding tools are added on the basis of HEVC: Intra Block Copy; Palette Mode (palette mode); Adaptive Colour Transform; Adaptive Motion Vector Resolution.

Compared with natural images, SC images have much fewer color types, that is, pixel values in a certain area tend to have greater repetition and similarity. With proper quantization, colors in a coding unit (CU) tend to appear as a few peaks. The palette mode is an encoder designed for this feature of SC. It analyzes and quantifies each pixel in the CU, selects several representative colors to generate a palette, and evaluates each color. assign a pointer. Each representative color is composed of three components, RGB or YUV. Individual colors that do not belong to representative colors after quantization are marked as escape pixels, and these pixel values and their pointers are directly encoded. The pointers in the palette are mapped back to the CU, the pixels in the CU are represented by pointer values to form a CU color pointer map, and the map is traversed, scanned and entropy encoded. However, the complexity of the palette mode is high, especially in the face of more complex color types, such as realistic games and animated movies, and the natural image parts in the mixed image, the generation and prediction of the palette will be more difficult.

Now, the video content of game live broadcast is becoming more and more popular, but no one has conducted experiments on the SC video content of games, so the present invention proposes a fast algorithm for game video content.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a HEVC-SCC adaptive energy consumption control strategy for game video content, according to the discrimination of a specific test sequence, to discriminate areas with complex colors (similar to natural images) and very flat areas, and The strategy algorithm to turn off palette mode for both regions.

A HEVC-SCC adaptive energy consumption control strategy for game video content of the present invention includes the following steps:

Step 1, input the video sequence image of the game to be processed;

Step 2, obtaining the brightness information of the game video sequence image image;

Step 3: Divide each frame image of the game video sequence into image blocks of 8×8 size on average;

Step 4, perform DCT transformation on each block to measure its color complexity, and obtain the absolute value a of the DCT coefficient;

Step 5, define two parameters i and k; i is a constant, which is a threshold for comparison with a, and k represents the flatness of a block;

Step 6, judge whether 0≤a≤i is satisfied, if 0≤a≤i, that is, when a is 0 or very small, execute step 7;

Step 7, the k value accumulates 1; if the absolute value a is greater than i, the k value remains unchanged;

Step 8: Calculate the upper threshold upK and the lower threshold lowK related to the average value aveK of all k values in one frame of image; the specific formula is as follows:

Among them, α and β represent two weight values, pixX, pixY respectively represent the number of horizontal pixels and vertical pixels of a frame of image, k ₁ +k ₂ +…+k _num represents all 8 in this frame The k value corresponding to the ×8 block, num represents the number of 8x8 blocks in one frame of image;

Step 9: Set the corresponding output value of a block to Jud _b , and all Jud _b form a discriminant matrix; the specific calculation formula of Jud _b is:

Compare k with upK and lowK: when the k value of an 8x8 block is greater than upK, the corresponding output value Jud _b of this block is 0; when the k value of a block is less than lowK, the corresponding output value Jud _b is 1;

Step 10: According to the size of the input CU, the elements Jud _b in the judgment matrix are fused: the fusion method is to perform a weighted average on all Jud _b ; the calculation formula of the flatness of the CU is as follows:

Among them, S _cu × S _cu represents the size of the CU, n represents that the CU has several 8x8 blocks, and m represents the number after fusion, that is, the flatness of the CU;

Step 11, judge whether 0<m<1 is satisfied;

In step 12, if 0<m<1 is satisfied, the default palette mode is enabled; if not, that is, when m is currently 1 or 0, step 13 is performed to close the palette mode.

Compared with the prior art, the HEVC-SCC adaptive energy consumption control strategy for game video content of the present invention can improve the coding rate for SC image coding and reduce coding time without affecting coding quality.

Description of drawings

FIG. 1 is an overall flow chart of the HEVC-SCC adaptive energy consumption control strategy for game video content according to the present invention.

Detailed ways

The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

The overall idea of the present invention determines the color complexity of the image, skips the palette mode for areas with high color complexity and extremely flat areas, and enables the default palette mode otherwise.

As shown in Figure 1, in a specific embodiment, various types of game video test sequences are collected, and the sequences collected in the first step are respectively subjected to an experiment of turning on and off the default palette mode and recording; the present invention includes the following steps :

Step 1, input the image to be processed;

Step 2, obtaining the brightness information of the image;

Step 3, divide the image into 8x8 blocks;

Step 4, perform DCT transformation on each block to measure its color complexity, and obtain the absolute value a of the DCT coefficient;

Step 5, define two parameters i and k; i is a constant, which is a threshold for comparison with a, and k represents the flatness of a block;

Step 6, judge whether 0≤a≤i is satisfied. If 0≤a≤i, that is, when a is 0 or very small, go to step 7,

Step 7, the k value is accumulated to 1; if the absolute value a is greater than i, the k value remains unchanged;

Step 8: Calculate the upper threshold upK and the lower threshold lowK related to the average value aveK of all k values in one frame of image; the specific formula is as follows:

Among them, α and β represent two weight values, pixX, pixY respectively represent the number of horizontal pixels and vertical pixels of a frame of image, k ₁ +k ₂ +…+k _num represents all 8 in this frame The k value corresponding to the ×8 block, num represents the number of 8x8 blocks in one frame of image;

Step 9: Set the corresponding output value of a block to Jud _b , and all Jud _b form a discriminant matrix; the specific calculation formula of Jud _b is:

Compare k with upK and lowK: when the k value of an 8x8 block is greater than upK, the corresponding output value Jud _b of this block is 0; when the k value of a block is less than lowK, the corresponding output value Jud _b is 1;

Step 10: According to the size of the input CU, the elements Jud _b in the judgment matrix are fused: the fusion method is to perform a weighted average on all Jud _b ; the calculation formula of the flatness of the CU is as follows:

Among them, S _cu × S _cu represents the size of the CU, n represents that the CU has several 8x8 blocks, and m represents the number after fusion, that is, the flatness of the CU;

Step 11, judge whether 0<m<1 is satisfied;

Step 12, if 0<m<1 is satisfied, the default palette mode is enabled; if not, that is, when the current m is 1 or 0 (as known from the above steps, m cannot be greater than 1 or less than 0), execute Step 13, turn off palette mode (skip palette mode).

Below is a description of the experiments. The encoder used in the experiment is HM-16.6+SCM-5.0, and the experimental platform is Intel Core i7 CPU-2.67GHz, 6G RAM. Experimental conditions use the all-intra-main-scc configuration file. QP is set to 22, 27, 32, 37. The constant value i is set to 8, aveK _up and aveK _low are set to 62 and 48, so the parameter thdK is 55. When aveK≥55, α is set to 0.7; when aveK<55, α is set to 0.75, and β is set to 0.1. Test sequences are shown in Tables 1 and 2. The test sequences in Table 1 are all game video sequences, and the test sequences in Table 2 are public test sequences. The experimental results are shown in Table 3 and Table 4.

Table 1. Game video sequences

Sequance name Resolution FPS YUV XXL 1920x1080 25 4:2:0 KYBC8CC 1280x720 25 4:2:0 WZRYZDX 1280x720 25 4:2:0 NBA2K17 1280x720 25 4:2:0 ChinaSpeed 1024x768 30 4:2:0

Table 2. Public test video sequences

Table 3. Experimental results of game video sequences

Table 4. Experimental results of public test video sequences

The BDPSNR gap of all test sequences is below 0.01dB. For the game test sequence set, the present invention has an average BDPSNR difference of 0.0259dB and a time saving of 15.715% compared with the default palette mode; for the public test sequence set, the BDPSNR average difference is 0.0402dB and a time saving of 8.319%. The experimental results are good.

The present invention performs a pre-judgment based on the video content before the palette mode, and skips the palette mode for the part of the video content that is more complex or close to a natural image, and reduces the coding time and time on the premise of ensuring the coding quality. required energy consumption.

Claims (1)

1. An HEVC-SCC adaptive energy consumption control strategy for game video content, comprising the following steps:

step (1), inputting video sequence images of a game to be processed;

step (2), acquiring brightness information of a game video sequence image;

step (3), dividing each frame of image of the game video sequence into 8 × 8 image blocks on average;

step (4), DCT transformation is carried out on each block to measure the color complexity of the block, and the absolute value a of a DCT coefficient is obtained;

step (5), defining two parameters i and k; i is a constant, is a threshold value, and is used for comparison with a, and k represents the flatness degree of a block;

step (6), judging whether a is more than or equal to 0 and less than or equal to i, and if a is more than or equal to 0 and less than or equal to i, executing step (7) when a is 0 or very small;

step (7), accumulating the k value by 1; if the absolute value a is larger than i, the k value is kept unchanged;

step (8), calculating an upper threshold upK and a lower threshold lowK related to an average value aveK of all k values in a frame image; the specific formula is as follows:

α and β represent two weight values, pixX and pixY represent the number of horizontal pixels and the number of vertical pixels of a frame of image respectively, and k₁+k₂+…+k_numRepresents the k value corresponding to all 8 × 8 blocks in the frame, and num represents the number of 8x8 blocks in a frame image;

step (9), setting the corresponding output value of one block to Jud_bAll of Jud_bForming a discrimination matrix; jud_bThe specific calculation formula of (A) is as follows:

compare k to upK, lowK: when the k value of an 8x8 block is greater than upK, the block corresponds to an output value of Jud_bIs 0; when the k value of a block is less than lowK, thenCorresponding output value Jud_bIs 1;

step (10) of fusing Jud elements in the discrimination matrix according to the size of the input CU_b: the fusion mode is to all Jud_bCarrying out weighted average; the flatness degree calculation formula of CU is as follows:

wherein S is_cu×S_cuIndicating the size of the CU, n indicating that the CU has several 8 × 8 blocks, m indicating the number after merging, i.e. the flatness of the CU;

step (11), judging whether m is more than 0 and less than 1;

step (12), if m is more than 0 and less than 1, starting a default color matching board mode; if not, namely when the current m is 1 or 0, executing the step (13) and closing the palette mode.

Images